Background

Many research questions in bacteriology, ranging from studying nutritional preferences (e.g., carbon substrates) to metabolic capacities or sensitivities against various compounds, including specialized metabolites from plants, antibiotics, or environmental chemicals, rely on chemical phenotyping, i.e., the quantitative assessment of bacterial growth during chemical exposure. When studying the plant root microbiome, chemical phenotyping is useful to dissect interactions of root microbes with plant root exudates. Root exudates are secreted to the surrounding soil and contain primary metabolites, on which microbes feed, as well as specialized metabolites with attractant and repellent properties [1–3]. These bioactive metabolites can affect bacterial behavior, such as motility, biofilm formation, and chemotaxis [4–8]. Studies on how specialized plant metabolites, such as coumarins or benzoxazinoids, contribute to the structure of root microbiomes are a prime area for chemical phenotyping [9–11].

Traditionally, bioactive compounds are tested using dilution methods, disc, gradient diffusion assays, or chromogenic media [12]. These methods are labor-intensive, have high material requirements, and need large amounts of compounds to be tested. Alternatively, commercially developed phenotyping arrays (e.g., www.biolog.com) permit screening and testing for metabolic capacities and other microbial properties. Such systems are based on 96-well plates, have much lower material requirements, and allow highly standardized growth and cross-referencing of results. One limitation is the restriction on the range of chemical compounds as offered by the company; thus, only broadly used and commercialized compounds can be tested. However, several specialized plant metabolites are not included in commercial phenotyping arrays or are not commercially available and need to be isolated and purified. A second limitation is linked to scalability and costs, particularly when a research question requires the chemical phenotyping of large sets of bacterial strains, such as established strain collections of the root microbiome [13–18]. In our research on how specialized plant metabolites affect the growth of root microbiome isolates, we sought to develop a custom approach satisfying the following requirements: (i) flexible and scalable, allowing simple variation in numbers of strains, compounds, and concentrations; (ii) adaptable to any dissolvable chemical compound, which is available in sufficient amounts and stable in the given conditions; (iii) high-throughput, where many reactions and replications can be run simultaneously for the production of robust and reproducible data; and finally, (iv) affordable for a small laboratory. Inspired by commercial solutions and previous work [19], we developed our custom approach for chemical phenotyping of bacteria fulfilling these requirements.

A few considerations regarding the necessary infrastructure need to be noted: plate readers are routinely used in molecular biology to measure absorbance, fluorescence, or luminescence in microtiter plates, and they are available in many laboratories. Plate readers are costly instruments (approximately 75,000€ in our case), and moderate investments can upgrade the scalability and throughput of experiments tremendously. The key investment is a stacker that automatically handles stacks of plates (typically between 15 and 25 plates; investment ~25,000€). A second investment is a benchtop 96-channel pipetting system (investment ~12,000€) that greatly speeds up the setup of many plates for replication of strains, compounds, and concentrations. The pipetting system makes it possible to set up a standard phenotyping run in less than 3 h, and the stacker then handles up to 25 × 96-well plates autonomously. Such a standard phenotyping run can include up to 60 strains in 3 replicates that are tested in 8 different conditions (compounds, concentrations, and controls). With hourly measurements, such a run will generate 70,000 datapoints in a 48-h assay. Overall, integrating high-throughput pipetting and automated measurements with multiple plates with a stacker significantly expands the capabilities of the plate reader, thereby enhancing the efficiency and scope of bacterial chemical phenotyping.

Here, we describe a customizable, high-throughput phenotyping approach that measures bacterial growth in the presence of chemical compounds originally developed in [20]. The basic requirements are that the bacteria grow aerobically, that the chemical compound of interest is stable in these conditions, and that, if a compound is not soluble in water, the used solvent does not prevent bacterial growth. Mild toxicity of the solvent can be controlled using appropriate controls. As the stacker and plate reader are not sterile environments, the assays require media controls (no-bacteria controls, NBCs) to detect potential contamination during a run. Using NBCs and media control plates, we typically do not detect environmental contamination for up to 68 h of assay length; however, contamination may depend on the media used. Our approach includes five steps, as illustrated in the Graphical overview. Preparatory steps are needed for the assay, including (i) pre-cultivation of the bacteria and (ii) preparation of the treatments, i.e., the stocks of the chemical compounds to be tested. Then, the 96-channel pipetting system helps to (iii) set up the assay plates, where the bacteria and their culture medium containing the different concentrations of the test chemicals are assembled. Then, (iv) plates are loaded onto the stacker, which handles them for autonomous recording of the optical density of the bacterial cultures over time with the plate reader. Finally, (v) exported data are analyzed using the open-source software R. We describe our approach based on an exemplary stacker run, for which we provide template data and a corresponding R script for analysis. In this example, we studied the antimicrobial activity of a specialized maize metabolite on the growth of maize root bacteria. Below, we provide the detailed protocol, including required materials and methodological steps. In the notes, we discuss limitations of our approach and offer recommendations for enhancing its scalability, as well as explore potential applications beyond the experimental example provided.